Disable constant folding (#26)

# Pull Request

### What change is being made

Disable constant folding

### Why this change is being made

Allow to export tasks on cuda

### Tested

Tests with task on cuda

GitOrigin-RevId: 0181833766b4b18c4d52cafaa9c729807362b9a6

Author: awoll-bdai

exporter/exporter/exporter.py

@@ -222,6 +222,10 @@ def export(
                 verbose=self.verbose,
                 input_names=input_names,
                 output_names=output_names,
+                # constant folding requires that all tensors are on the same device. as constants
+                # are on cpu, we disable it to allow exporting of models on cuda. constant folding
+                # optimization will be done in the ONNX runtime when we load the model for deployment.
+                do_constant_folding=False,
             )
 
         wrapper_model = construct_decimation_wrapper(